Waitman Gobble wrote:
> Amos Jeffries wrote:
>> ?huh? your example omits cache_peer order.
>> Which is half the critical config for failover sequencing.
>> {important stuff in previous email}
>> Amos
>>
ok this squid.conf seems to be working for me. sans cache-mgr stuff.
http_port 80 accel defaultsite=waitman.net vhost
access_log /usr/local/squid/var/logs/access.log squid
cache_store_log none
cache_mgr {email}
cachemgr_passwd {assword} all
cache_dir ufs /cache 307200 16 256
maximum_object_size 65536 KB
cache_mem 1208 MB
cache_effective_user squid
cache_effective_group squid
coredump_dir /cache
acl s1 dstdomain "/usr/local/squid/etc/s1"
http_access allow s1
acl s2 dstdomain "/usr/local/squid/etc/s2"
http_access allow s2
http_access deny all
cache_peer 172.16.10.5 parent 80 0 no-query originserver name=server_s1
login=PASS
cache_peer_access server_s1 allow s1
cache_peer 172.16.10.7 parent 80 0 no-query originserver name=server_s2
login=PASS
cache_peer_access server_s2 allow s2
cache_peer 172.16.10.5 parent 80 0 no-query originserver name=bu_s1
login=PASS
cache_peer_access bu_s1 allow s2
cache_peer 172.16.10.7 parent 80 0 no-query originserver name=bu_s2
login=PASS
cache_peer_access bu_s2 allow s1
Basically I wanted an inexpensive solution for redundancy, hosting many
sites on inexpensive equipment. The current test: With four servers, 2
squid machines and 2 main servers - if one goes down everything still
works. If a squid machine and a main server goes down at the same time
then everything still works. If both squid machines or both main servers
go down, then I'm SOL but that's unlikely... They're spread out
geographically so it would 'prolly take something like a nuclear war for
that to happen. (crossing fingers).
The squid machines are also ns servers, which say 'i'm master of domain
x, i serve http for domain x' - the domain in question. So the
redundancy begins at the nameserver field in the domain record.
So it seems to work so far with my limited test, I'll see what happens
if I set up a bunch of domains. It also seems scalable.
However, you mentioned "cache_peer order" - it seems to go in the order
specified in squid.conf - i've reviewed the docs - i don't see a
parameter to set the order?? just checking. maybe i missed something.
Waitman
Received on Mon Aug 24 2009 - 05:59:22 MDT
This archive was generated by hypermail 2.2.0 : Mon Aug 24 2009 - 12:00:04 MDT